Episodic anime stories
The same characters across episodes and scenes.
DomoAI · Multi-reference video
DomoAI Omni Reference
Guide videos with images, clips and audio in one generation
Upload the character, the location, the move and the sound, then tell DomoAI what each reference is for. Omni Reference combines them into one video, so characters stay consistent from shot to shot.
50references max
4–30 svideo length
2models
One generation, many references
IMG
Image 1First frame & location
IMG
Image 2Character identity only
VID
Video 1Movement & camera rhythm
AUD
Audio 1Dialogue or soundtrack
One consistent videoUp to 30 s · 480P–2K depending on model
The basics
What is Omni Reference?
Omni Reference is a separate creation tool in DomoAI, launched in August 2026. It builds one video from many references at once: images, video clips and audio files. Image to Video animates a single picture. Omni Reference lets you give each reference a job, such as who the character is, where the scene happens, how it moves and what it sounds like.
- Keep a character’s face, hair and outfit consistent across shots
- Borrow motion and camera rhythm from a reference clip
- Drive dialogue or music timing with an audio file
- Generate even from audio alone with Seedance 2.5
30images
10video clips
10audio files
Maximums on Seedance 2.5. Video and audio clips are 2–30 s each, with up to 30 s total of each type.
Models
Seedance 2.5 vs MiniMax H3
Choose based on whether you need more references and longer clips, or higher resolution.
| Spec | Seedance 2.5 (ByteDance) | MiniMax H3 (MiniMax) |
|---|---|---|
| Images | Up to 30 | Up to 9 |
| Video clips | Up to 10 | Up to 3 |
| Audio files | Up to 10 | Up to 3 |
| Output length | 4–30 s | 4–15 s |
| Resolution | 480P or 720P | 768P up to 2K |
| Best for | Complex scenes, longer clips, audio-led videos | Sharper, shorter hero shots |
Neither model outputs 4K directly. Use the Video Upscaler afterwards. Specs are from DomoAI’s Omni Reference page; check the tool for the latest limits.
Reference roles
Give every reference one job
References work best when your prompt says what each one controls and what it must not change. A reference without a role adds confusion, not control.
| Reference | Its job | Must not |
|---|---|---|
| Image 1 | First frame, location, camera position, light direction | — |
| Image 2 | Character’s face, hairstyle and wardrobe | Change framing, composition or lighting |
| Image 3 | One prop or space | Change the character’s identity |
| Video 1 | Movement timing and camera rhythm | Change identity or wardrobe |
| Audio 1 | Soundtrack or dialogue timing | Drive the cut timing |
Use far fewer than 50. Two or three images with clear jobs usually beat a pile of unlabeled ones.
Interactive
Omni Reference prompt builder
List your references, give each one a role, and get a prompt that tells the AI what each one controls. The builder warns you about model limits as you go.
Model
References
Your prompt
Runs in your browser. The exact credit cost shows on DomoAI’s Generate button before you spend anything.
Step by step
Beginner’s guide →How to use Omni Reference
- 01
Open Omni Reference
Sign in to DomoAI and open Omni Reference from the create menu.
- 02
Upload references
Add images, video clips (2–30 s) and audio files (2–30 s). Upload only what has a job.
- 03
Assign roles in the prompt
For example: “Image 1 is the first frame. Image 2 controls the character’s face and outfit only.”
- 04
Choose settings
Model, duration, aspect ratio, resolution and audio on or off.
- 05
Check the cost & generate
The credit cost appears on the Generate button. Start short and at 480P to test.
- 06
Polish
Upscale, add lip sync or extend the story with the next shot.
Consistent characters
The anchor-image method
DomoAI’s own advice: “The fix is an anchor image, not a better sentence.” A longer prompt won’t stop a character’s face from changing between shots. A good reference image will.
- Make one anchor: front-facing, full body, plain background
- Make variants from the anchor, never from another variant, so changes don’t build up
- Lock five things: face, hair, body scale, outfit, voice
- Describe each locked trait once, either in a reference or in the prompt, not both
The clause to add
“Image 1 controls the location, camera position, light direction and composition, and is the literal first frame. Image 2 controls the character’s facial identity, hairstyle and wardrobe ONLY. Do not recreate its framing, composition, or lighting.”
Test in this order
- Static shotDoes the face hold?
- Motion shotDoes it hold while moving?
- Location changeDoes it hold in a new place?
If the character still looks the same after all three, longer sequences usually stay consistent too.
Troubleshooting
Common problems and fixes
| Problem | Real cause | Fix |
|---|---|---|
| Face drifts mid-clip | Only one anchor point | Add a second anchor, such as a first frame and a last frame |
| Reference hijacks the shot | No exclusion clause | Add “do not recreate its framing, composition, or lighting” |
| Outfit changes between shots | Outfit described in both the reference and the prompt | Keep one source of truth and delete the other |
| Garbled text on signs or clothes | AI-generated text | Remove the text and add it in editing |
| Random results | Too many references without roles | Cut down to the references that have a job |
Which tool?
Omni Reference vs other DomoAI tools
| Tool | Input | Best for |
|---|---|---|
| Omni Reference | Many images + clips + audio, each with a role | Consistent characters, multi-shot stories, music videos |
| Image to Video | One image + prompt | Quickly animating a single picture |
| Motion & Style Reference | Character + motion clip + style image | Copying a specific move or look |
| Frames to Video | Start and end frames | Controlled transitions between two shots |
| Character to Video | A character + scene prompt | Placing a character into new scenes |
Use cases
All use cases →What people make with Omni Reference
Music videos
The track as an audio reference, with characters performing to it.
Dialogue scenes
The voice line sets the timing; the image keeps the speaker’s identity.
Product & brand ads
The product, mascot and location stay the same across every cut.
Limitations & costs
- No direct 4K: 720P max on Seedance 2.5, up to 2K on MiniMax H3
- Clips are 4–30 s, so longer stories are built shot by shot
- Credit cost rises with more uploads, longer clips and higher resolution
- Generated text on signs and clothing is unreliable
- New accounts get 30 credits, enough for one 4 s generation at 480P
Use references responsibly
Upload only images, clips and audio that you own or have permission to use. Don’t use a real person’s face or voice without consent, and never create misleading deepfakes. Label AI-generated content where platforms require it.
Terms & usage policy →What is DomoAI Omni Reference?
A DomoAI creation tool that builds one video from many references at once: images, video clips and audio files. You assign each reference a role in your prompt, such as the character, the first frame, the motion or the soundtrack.
How many references can I upload?
With Seedance 2.5, up to 50: 30 images, 10 video clips and 10 audio files. Video and audio clips are 2–30 seconds each, with up to 30 seconds total of each type. MiniMax H3 takes up to 9 images, 3 video clips and 3 audio files.
Which models does Omni Reference use?
Seedance 2.5 (4–30 s, 480P or 720P) and MiniMax H3 (4–15 s, 768P up to 2K). Neither outputs 4K directly. Use the Video Upscaler afterwards.
Do I have to assign a role to each reference?
It isn’t required, but it makes results far more predictable. Say what each reference controls — for example, “Image 2 controls the character’s face, hair and outfit ONLY” — and what it must not change.
Can I generate a video from audio only?
Yes. DomoAI says audio-only generation is possible with Seedance 2.5.
How is Omni Reference different from Image to Video?
Image to Video animates one picture. Omni Reference combines many images, clips and audio files, each with an assigned role, which makes it better for consistent characters and multi-shot stories.
How much does Omni Reference cost?
It depends on what you upload, how long the clip is and the resolution you pick. The exact credit cost appears on the Generate button before you spend anything. New accounts get 30 credits, enough for one 4-second generation at 480P.
How do I keep my character consistent?
Use an anchor image: a clear, front-facing, full-body shot. Assign it the identity role with an exclusion clause, and describe each trait only once. Test with a static shot, a motion shot and a location change before making longer videos.
Show it, don’t just say it.
Give DomoAI your character, your move and your sound, and get one video that keeps them all consistent.